You figured out AI. Your team hasn't.


Hello Reader

Most marketing leaders don't fully grasp the work required to build an AI-ready marketing function.

I see it in workshops, advisory calls, and the team audits I run with clients.

The conversation usually goes like this:

We've been experimenting with AI, we've got a few people using tools, and when I ask what's on their workflow map, there's a long pause.

They don’t have an answer.

Salesforce's 2026 State of Marketing report, which surveyed 4,450 marketers, found the same pattern, with 61% of marketing teams saying their AI adoption is high but their workflow integration is still a work in progress.

The problem is that nobody has mapped where AI belongs in the sequence of work, and until that map exists, AI stays in the hands of your superusers, and out of reach for everyone else.

If you’re suffering from random acts of AI, widening skill gaps and bottlenecks that haven't moved, a workflow mapping session is where you start.

This edition walks you through how to run one.

Why the tools aren't the problem

An INSEAD and Harvard Business School study tracked 515 startups through a structured accelerator program. Both groups had access to the same AI tools, but one group used AI to speed up individual tasks and the other was shown how to reorganize their operations around AI.

The workflow redesign group generated 1.9x higher revenue, were 18% more likely to acquire paying customers and completed 12% more tasks compared to the group who applied AI to the same workflows.

McKinsey's 2025 State of AI survey confirmed the same pattern at enterprise scale — organizations reporting significant financial returns from AI were twice as likely to have redesigned their workflows before choosing their tools.

What mapping actually means

You can't redesign a workflow you haven't mapped, so before any AI decision, document every step in the workflow you're redesigning.

As an example, your content creation workflow isn't "create content."

It's researching topics, writing briefs, drafting, editing, adapting for each channel, scheduling and distributing. Each of those is a discrete step with its own logic, owner and AI potential.

Once every step is mapped, lay them out in order.

This is usually where you’ll discover the hidden handoffs and bottlenecks.

MIT Sloan research also found that how tasks are clustered matters just as much as which tasks you hand to AI.

With your sequence visible, you can remove those bottlenecks and unnecessary hand offs. You might even be able to connect multiple steps into a single chain you delegate entirely to AI.

How to run a workflow mapping session

Collect everyone who contributes to the workflow you're mapping. Bring the full cast, whatever the number.

Block 60 to 90 minutes and commit to documenting every step in detail.

A meeting room with a whiteboard is ideal, but if you're working with a distributed team, Miro or a similar virtual whiteboard works well.

Pick one workflow to trace.

The best candidate is something the team ran in the last 90 days.

It could be a campaign, a content series, or a product launch. Walk through every step, from brief to publish, and write each one down as you go.

At each step, ask three questions:

#1. Does this require human judgment, or is it pattern-based work that repeats the same logic every time? Judgment stays with your team, pattern-based work is a candidate for AI.

#2. How often does this step happen? Daily and weekly tasks are your first priority. Quarterly and annual ones are worth noting but not where you start.

#3. What's the cost of getting it wrong? A wrong first draft costs 30 minutes. A wrong campaign message costs you pipeline. This determines your human oversight.

There's a secondary benefit to this process that most leaders don't anticipate.

Team members who've been fuzzy on how their work connects to everyone else's leave with a clear picture of the full sequence, and when a team redesigns a workflow together rather than having a new process handed to them, are far more invested in making it stick.

Pick a Lane for Every Step

Before you assign AI to any workflow step, you need to decide how much oversight that step requires.

Assistive. Your team member does the work collaboratively with AI, applying their judgment and guiding the AI every step of the way.

Human-in-the-loop. The team member kicks off the task and reviews the output at defined checkpoints, but the AI handles the execution. This is the right model for high-frequency, lower-stakes work.

Autonomous. The agent runs the full workflow end to end with no human touch. Reserve this for tasks that are repetitive, well-defined and low-risk.

If you're not sure which mode fits a given step, start with human-in-the-loop. You'll catch where the AI breaks before it causes a problem, and your team builds the judgment to extend AI's role from a position of experience.

People, process, data

Once you start redesigning your workflows, you’ll find it starts impacting other areas of your function, including your people, processes and data management.

Here’s what you need to consider for each:

People.

Before any workflow goes live with AI, your team needs to understand what's changing and why. That means being explicit about which steps are moving to AI, what the human role looks like inside those steps going forward and what skills people need to work effectively alongside the AI.

Process.

AI initiatives need to be tied to specific business outcomes to deliver measurable value. An AI agent built to reduce campaign turnaround time by 40% is a process investment with a success metric. An agent built because a vendor gave a compelling demo is an experiment with no finish line. Every workflow you redesign should map back to a specific goal your team is trying to move.

Data.

Your AI agents are only as good as the data powering them. Before you build anything, make sure the data they need is clean, accessible and in one place.

What happens when AI takes the task?

Most workflow mapping sessions end with a build list and a lot of enthusiasm. What they rarely address is what happens to the person who owned that work.

Before any pilot goes live, the owner of each automated step should be able to answer three questions:

What am I no longer doing?

Identify the tasks moving to AI and the hours per week that frees up.

What am I doing instead?

Identify the higher-judgment work that's been deprioritized because the repetitive work consumed too much time. Good candidates for this are the analysis project that never got done, the briefs that got rushed, or stakeholder relationships that previously were neglected.

What do I need to learn?

Working alongside an AI agent requires a different set of skills than doing the task manually. As an example, reviewing AI output critically, writing effective prompts and knowing when to override the AI.

The Workflow Mapping Starter Kit

Here’s a reference guide you can use in your next session.

Step 1: Pick Your Workflow

Score each workflow 1-3 on these criteria:

Highest score = where you start. Seven or above is a strong first candidate.

Step 2: Map It With Your Team

60-90 minutes using a whiteboard or Miro.

Trace one workflow from the last 90 days. Then ask:

  • What's missing?
  • Where does it stall?
  • Where does it break?

Step 3: Assign a Lane to Every Step

  • Assistive: human and AI working together
  • Human-in-the-loop: AI executes, human reviews
  • Autonomous: AI runs end to end

Not sure? Start with human-in-the-loop.

Step 4: Build and Track

  • Pick your top three pilots from the highest scoring workflows
  • Assign one owner per pilot
  • Set a 30-day checkpoint

Want to Level Up Your AI Game?

If your team is ready for a hands-on AI strategy session, my custom-designed workshops are built to uncover the workflows that can save you hours every week.

Prefer to start small? My YouTube channel is packed with quick, practical “how-to” videos that show you exactly how I use AI tools for marketing, content, and automation.

Planning an event or conference? I deliver high-energy AI sessions that engage audiences and leave them with actionable strategies they’ll talk about long after the event. Book me for your event here.

Did some one forward you this email? You can subscribe here.

2120 Contra Costa Blvd #1059 , Pleasant Hill, CA 94523
Unsubscribe · Preferences

AI at Work

AI at Work is a weekly newsletter on how marketing teams redesign workflows, roles, and systems with AI. Real examples, practical frameworks, and repeatable processes operators can use immediately. Join thousands of successful marketing leaders by subscribing below!

Read more from AI at Work

Hello Reader Claude was a long-time cult favorite, beloved by developers and largely invisible to everyone else. Then, late last year, it swept across the AI conversation and is now one of the most popular AI platforms in the world, regularly trading places with ChatGPT at the top of the charts. People and teams love it. And almost everyone I talk to is confused by it. That confusion has a specific cause: Claude was originally designed for coders, and much of the language and systems are...

Hello Reader One of my favorite moments from the past month happened on a call with Charlie Treadwell, CMO at Elisity, and Jake Milstein, VP at Contrast Security. Charlie was telling us about the day one of his sales reps came to him asking if she could have "the PowerPoint template that looks like the new ones Claude is creating." He told her: "There is no template. You should never use a template ever again. Just tell Claude to make you the slide you want." She had a lightbulb moment and...

AI Content Creation

Hello Reader Years ago, when I was VP of Marketing at a SaaS platform, I had a content problem I couldn't solve with budget. We were too small to justify a full-time writer and I was already doing the work of three people, so doing it myself wasn't realistic. The answer was freelancers. I'd find a writer, brief them on our product and audience (corporate travel managers, a niche with its own language and priorities) and wait two weeks for something I could publish. What came back was almost...